Rating calculation


Rating calculation

Back to forum

Pablo Schmid    (2009-12-26)
Rating calculation

Hello Thibault, j'aurais préféré parler en français mais puisque je suis sur le forum...

Could you explain how exactly chess rating's calculation works in ficgs?

And I think games with 10 moves or less should be counted, maybe not every games but sometimes 8 or 9 moves can be sufficient to punish someone (nice miniatures or quick mates are possibles) and it enables the loser to resign (or be mated) without losing points. An example of an abusive situation: imagine I play the Traxler and I resign before 10 moves if I see a good choosed line by my opponent which I don't like to play as Black because I know I might suffer or even lose...

J'espère que je t'aurai convaincu et que mon exemple ne donnera pas de mauvaises idées! Sur ce, bonnes fêtes, Pablo

Thibault de Vassal    (2009-12-27 01:50:07)
Chess rating calculation

Hello Pablo! Bonnes fêtes également :)

All rating rules are explained here (there's a french version), feel free to ask if you have any question.


About the 10 moves rule, why 10 more than 9 or 8... well there must be a clear rule & easy to remember. I don't think that miniatures are a big problem, there may be an "unfair" situation in some rare cases, but it shouldn't happen at 2100+ elo ratings. The rule may be not perfect but I did not find a better one yet (any idea? :)).

Pablo Schmid    (2009-12-27 02:02:01)
rules of 10 moves

Thanks for the quick response, my proposition would be no limit of move at all to win points, as in OTB chess. Maybe an idea could be to not make winning points in a game where the player did not connect for a long time before the tournament begin as it is clear that it is a "forfeit", as in OTB when someone don't come.

But maybe you will convince me that your idea is better than mine?

Pablo Schmid    (2009-12-27 13:42:31)
Rating calculation

Salut Thibault, bon j'écris en français cette fois, j'ai lu les explications sur les calculs mais ça me semble assez obscur, alors je vais poser des questions proches de ma situation. Imaginons que je commence un tournoi 2000-2200 avec un rating à 2000 pile et que je fais un score de 50% contre des joueurs à 2050 de moyenne, en théorie je devrais gagner quelques points non? Et maintenant mon classement réel a monté depuis le début du tournoi disons à 2120, et du coup le 50% contre cette moyenne à 2050 me ferait perdre des points si on compte mon rating le plus récent. Alors est-ce que c'est mon ELO de départ qui va être pris en compte ou mon ELO actuel? En gros je suis en train de me demander s'il est intéressant de commencer un tournoi avant la nouvelle liste avec un classement supérieur sur cette liste, ou si cela ne change rien voire désavantageux...

J'espère que tu auras compris ma question et merci d'avance pour ta réponse :)

A +


Thibault de Vassal    (2009-12-27 14:08:37)
10 moves rule

IMO this rule is important because :

1) It dissuades cheating by creating several accounts playing together through proxies... This rule makes it really hard to win some points this way, it would be detected even more easily.

2) In many cases, new players (who did not realize that computer assistance was authorized, who do not like the correspondence time controls or who just wanted to try) forfeit their games after 3 or 4 moves rather than let it go. There is no doubt to me that this phenomenon would have much more bad effects on ratings.

3) It is likely that a player who "miniatures" another player is actually much stronger than his opponent, so his rating shouldn't increase so much.

Finally and that's the main point IMO, "unfair" situations are statistically negligible compared to the other possible rules. See the other servers...

So far I'm quite convinced that it is one of the best implemented rules here, and this is exactly the way I optimise the programs: "Statistics give better results than looking for perfection" :)

Thibault de Vassal    (2009-12-27 19:13:04)
Rating calculation

Re... nos réponses précédentes se sont croisées, désolé :) Le prochain classement est calculé en tenant compte du classement courant donc cela ne change rien puisque l'on ne peut pas vraiment prévoir quand se termineront les parties... on gagne forcément des points en faisant 50% contre des joueurs qui avaient un TER plus fort que le classement courant, idem dans l'autre sens. En général, la "gestion" des résultats n'influe que très peu sur le classement final.